brawer / osmdiffs / 35114827863
93%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 52
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 92.742%. Remained the same
35114827863

push

github

web-flow 
docs: add learnings from a real Infomaniak Kubernetes run (#817)

* docs: add learnings from a real Infomaniak Kubernetes run

A first production-shaped run of v0.8.5 against the real planet, on
Infomaniak Public Cloud Kubernetes (brawer/production's infomaniak/ +
infomaniak-k8s/) - not a permanent deployment (cluster torn down right
after), but real enough to surface things this document didn't cover:

- Real total runtime (5h55m) vs. this document's Hetzner-derived
  ~2h43m, with three ranked, explicitly unconfirmed hypotheses for the
  gap (entry-level compute tier, network-attached volume, the two
  compounding under this run's memory pressure).
- The page-cache-vs-heap design (see "What to watch") held under real
  memory pressure - cited with actual rss_anon_bytes/cgroup numbers.
- render_conflated_detail and join_conflated_tiles, not discussed
  elsewhere in this document, completed fine and needed no tuning.
- Three Kubernetes-specific operational gotchas with no equivalent in
  the bare podman invocation this document documents: fsGroup for a
  CSI-provisioned volume (vs. this doc's host-directory chown),
  active_deadline_seconds sizing (a Hetzner-derived guess killed a
  still-working run), and the AWS SDK's client-side region-string
  validation (lowercase-only, caught only at the very last step).
- The cost model differs by more than the hourly rate: an
  always-on-node platform bug forced billing-per-hour-regardless
  rather than this document's per-run Hetzner model.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Jh5d3mRja4ZGL6UBFA76ZR

* docs: add Infomaniak per-run cost, next to the Hetzner figure

One data point (the run this PR's "A first real run" section
describes) rather than a sweep, but directly comparable to the
Hetzner per-run figure above it: ~CHF 0.40 (compute + the ephemeral
scratch volume, at rates read straight from the account's own API),
same order of magnitude... (continued)

10490 of 11311 relevant lines covered (92.74%)

1369.91 hits per line

Jobs
ID Job ID Ran Files Coverage
1 35114827863.1 52
92.74
 GitHub Action Run
Source Files on build 35114827863